Copyright © IFAC Artificial Intelligence in Real-Time Control, Delft, The Netherlands, 1992 RIGAS: AN EXPERT SERVER TASK IN REAL-TIME ENVIRONMENTS* A. Crespo, J.L. Navarro, R. Vivo, A. Garcia and A. Espinosa Grupo de AUlomalica e Informalica IndUSlria/, Universidad Polilecnica de Valencia, Apdo 22012, £46071 Valencia, Spain Abstract. The integration of expert systems in real time environments provides new pos- sibilities to the process control of complex systems. Intelligent tasks working together with traditional control tasks under real time constraints introduce some problems in order to guarantee a response time of the system. In this paper, the use of a a real time expert sys- tem for process control, called RIGAS, encapsulated in an expert server task is described. The expert server is formed by an scheduler, a blackboard, and a set of task with real time constraints. Keywords. real time expert systems, intelligent control, blackboard system INTRODUCTION Real time and expert systems provide a general framework to solve a great number of problems not completely solved by means of traditional ap- proaches. One of the most interesting fields to apply these techniques is the control of complex systems. There is a significant number of topics in research re- lated with this subject. Some of them are: real-time scheduling, temporal representation and reasoning, guaranteed response time, uncertainty management, efficient inference, etc. [Laffey88] Blackboard systems are being used to experiment knowledge based systems in real time frameworks and it has been seen as one of the best alternatives to match complexity and variety of actual applications. However, expert system need to be integrated with other software components to complete the overall system including critical and non critical tasks. One of the main problem emerged is the predictable be- havior of the tasks with AI techniques. Most of the real time expert system are based in blackboard architectures [Hayes90] [Erickson90] • Partially supported by a grant ofCICYT No. ROB89- 0442 of Spanish Government 497 [PauI91] [REAKT90]. Real time expert systems knowledge bases use the ob- ject oriented approach to represent the knowledge. It allows an easy understanding, redundancy avoidance, and modularity, among other interesting issues. The features above stated require: • Guaranteed response time When we combine process control with expert systems techniques and we try to apply them under real time con- straints, some criteria have to be applied in or- der to guarantee the response time of the sys- tem. This guaranteed response time may be obtained at different levels: 1. Scheduling: establishing a plan to execute the appropriated tasks. The appropri- ated with respect to the deadline speci- fication and the resources used. 2. Knowledge Base: Rules are defined at dif- ferent levels of knowledge (deep knowl- edge), and several sets of rules cooperate in a problem. The inference engine knows which rules are appropriated and the time they require in order to solve a subprob- lem.